Model Calibration

A concept measuring the degree to which a model's output confidence scores align with its actual accuracy. A well-calibrated model produces predicted confidence levels that closely match empirical correctness rates, making calibration an important indicator of AI system reliability.